Hydrophilic membrane-based humidity control

Paul Scovazzoa,*, Jedrick Burgosa, Alex Hoehnb, Paul Todda

aDepartment of Chemical Engineering, University of Colorado, Campus Box 424, Boulder, CO 80309-0424, USAbBioServe Space Technologies, University of Colorado, Campus Box 429, Boulder, CO 80309-0429, USA

Accepted 30 April 1998

Abstract

A dehumidi®cation system for low gravity plant growth experiments requires the generation of no free-liquid condensate

and the recovery of water for reuse. In the systems discussed in this paper, the membrane is a barrier between the humid air

phase and a liquid-coolant water phase. The coolant water temperature combined with a transmembrane pressure differential

establishes a water ¯ux from the humid air into the coolant water. Building on the work of others, we directly compared

different hydrophilic membranes for humidity control. In a direct comparison of the hydrophilic membranes, hollow ®ber

cellulose ester membranes were superior to metal and ceramic membranes in the categories of condensation ¯ux per surface

area, ease of start-up, and stability. However, cellulose ester membranes were inferior to metal membranes in one signi®cant

category, durability. Dehumidi®cation systems using mixed cellulose ester membranes failed after operational times of only

hours to days. We propose that the ratio of ¯uid surface area to membrane material area (�membrane porosity) controls the

relative performances among membranes. In addition, we clari®ed design equations for operational parameters such as the

transmembrane pressure differential. This technology has several potential bene®ts related to earth environmental issues

including the minimization of airborne pathogen release and higher energy ef®ciency in air conditioning equipment. Utilizing

these study results, we designed, constructed, and ¯ew on the space shuttle missions a membrane-based dehumidi®cation

system for a plant growth chamber. # 1998 Elsevier Science B.V. All rights reserved.

Keywords: Membrane condenser; Dehumidi®cation; Theory; Microporous and porous membranes

1. Introduction

The design of a plant growth chamber for space

¯ight, `̀ PGBA'', drove this work on humidity control

[1]. The PGBA is a space experiment payload

designed to grow vascular plants in low gravity within

a well-controlled environment chamber capable of

maintaining approximately 50 cultivated plants in

an area of 0.1 m2 and a volume of 0.03 m3 (30 l).

This ratio of surface to volume coupled with plant

transpiration rates places a large burden on the cham-

ber's humidity control system, which must do the

following:

1. Operate in low gravity.

2. Generate no free-liquid condensate in contact with

the humid air being treated, for the protection of

electronic equipment and the limitation of patho-

gen growth.

3. Recover water vapor condensate for reuse by the

plants.

A number of other researchers have proposed and

used membrane-based dehumidi®cation systems. The

Journal of Membrane Science 149 (1998) 69±81

*Corresponding author. Fax: +1-303-492-7471.

0376-7388/98/$ ± see front matter # 1998 Elsevier Science B.V. All rights reserved.

P I I : S 0 3 7 6 - 7 3 8 8 ( 9 8 ) 0 0 1 7 6 - 8

work of these researchers at the Wisconsin Center for

Space Automation and Robotics (WCSAR) [2], Bend

Research [3], and Lockheed [4] con®rmed the premise

that cooled porous membranes can control humidity.

Each of these research groups tested a different mem-

brane material and different operational principles. In

the present study, instead, we directly compare the

performance of different membrane materials and

rigorously de®ne the operational principles for hydro-

philic membrane dehumidi®cation systems. We

believe that the comparisons and operational principle

de®nitions contained within this paper are necessary

for the further and ef®cient progress of research into

membrane-based dehumidi®cation. In this paper we

compare the performance of the following hydrophilic

membrane materials; the research group indicated

after each material indicates whose work we used

as background for our continued work with said

material:

� Sintered metal (WCSAR, US Patent 5 368 786)

[2].

� Ceramic (BioServe Space Technologies) (this

study).

� Mixed cellulose ester (Bend Research) [3].

2. Theoretical development

2.1. Dehumidification background

Before proceeding, it is prudent to review the

important theoretical principles of dehumidi®cation.

Fig. 1(a) shows the basics of conventional earth-based

humidity control systems that operate below the dew-

point temperature. The schematic psychrometric chart

on the right tracks the properties of water concentra-

tion and temperature as the humidity content of air

moves from point 1 to point 2. The physical repre-

sentation on the left shows how a cold ¯uid (refrig-

erant) passing on one side of a solid condensing

surface (metal) moves the humid air through the

pathway shown on the psychrometric chart and gen-

erates a liquid condensate in contact with the humid

air being treated.

Fig. 1(b) shows what will happen if a water vapor

concentration gradient is setup across a material

permeable to water vapor. In this case, the air dehu-

midi®es via diffusion without a change in bulk air

temperature. Such systems exist; however, they oper-

ate at high pressures inappropriate for integration into

Fig. 1. Conventional dehumidification systems. (1a) Dew-point dehumidification (condensing heat exchanger) showing, on the left, the

physical operation and on the right, a schematic of the psychrometric cycle. (1b) Diffusion-based dehumidification, such as a water vapor

selective dense film membrane system operated under cross membrane pressure, �P.

70 P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81

space-¯ight experiment hardware. Fig. 2 shows the

result of combining the two processes of Fig. 1 by

using a porous membrane. The low vapor pressure of

cold liquid water establishes a concentration gradient

for transporting water vapor out of the humid air and

across the membrane into the cold liquid water. Since

there is also a temperature gradient, the simultaneous

transport of heat alters the bulk air temperature. The

heat transfer is only an artifact of using cold liquid

water; and the bulk humid air does not cool to its dew

point as is the case in the conventional, dew-point

temperature, earth-based system (Fig. 1(a)).

The combined process in Fig. 2 has three distinct

advantages over the conventional dew-point tempera-

ture system: ®rst, no free-liquid condensate is in

contact with the treated humid air; second, direct

reclamation of the condensate as liquid water; and

third, a higher theoretical thermal ef®ciency. It is this

combined scenario that the researchers mentioned in

Section 1 have applied.

2.2. Membrane-based dehumidification

Simply stated, in membrane-based dehumidi®ca-

tion, the membrane is a barrier between the humid air

and liquid water with the vapor pressure of the liquid

water (a function of temperature) establishing a water

¯ux. In terms of humidity, the molar water ¯ux is

j � ko�loc�P

RT

� �Ha=18

1=29�Ha=18ÿ Hw=18

1=29�Hw=18

� �;

(1)

where j is the molar water ¯ux (mol/m2/s), ko(loc) the

local overall mass transfer coef®cient (m/s), P the air

pressure (Pa), R the ideal gas constant (m3 Pa/mol/K),

T the absolute temperature (K), Ha the absolute

humidity of the bulk air (kg of H2O/kg of dry air),

Hw the absolute humidity of air in equilibrium with

the liquid water (kg/kg), 29 the molecular weight of air

(kg/kmol), and 18 is the molecular weight of water

(kg/kmol). The ratios in the brackets are mole frac-

tions.

The existing literature [5] presents the following

model for the mass transfer coef®cient in a hydrophilic

membrane dehumidi®cation:

1=ko � 1=kbl � 1=km; (2)

where ko is the overall mass transfer coef®cient (m/s),

kbl the air/membrane boundary layer mass transfer

coef®cient (m/s), and km is the membrane mass trans-

fer coef®cient (m/s), with km being a function of

Darcy's law for liquid ¯ow in porous media [5]. We

do not agree with the dependence of km on Darcy's law

in this application as re¯ected in Eq. (2); below we

develop an alternative model for ko as presented in

Eq. (5).

2.3. Transmembrane pressure

The existing literature model makes the membrane

mass transfer coef®cient a function of transmembrane

pressure through Darcy's law for liquid ¯ow through

saturated porous media that obeys Darcy's law

J � K��P�=X; (3)

where J is the water ¯ux, K the membrane perme-

ability, �P the dynamic ¯uid pressure difference

across the membrane, and X is the wetted membrane

Fig. 2. Two-phase membrane-based dehumidification. Physically, on the left, the membrane separates the humid air phase from the liquid cold

water phase. The driving force for mass transport is the difference between the water vapor pressure in the humid air and the vapor pressure of

water in equilibrium with the cold liquid water. On the right, the psychrometric cycles indicate the effects of mass transport and heat transfer,

an artifact of using cold liquid water to establish the water vapor concentration gradient.

P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81 71

thickness. This pressure differential results from oper-

ating the cooling water at a pressure below the ambient

air pressure, which is assumed to be the pressure of the

humid air stream.

We did not ®nd the mass transfer coef®cient to be a

function of the transmembrane pressure, �P. The

transmembrane pressure did not have any effect on

condensation rate; therefore, we concluded that Darcy

¯ow through hydrophilic membranes was not a con-

densation-rate-determining step. While we do not

reject Darcy ¯ow for ¯uid transport through the

membrane, we propose that any transmembrane pres-

sure between two ®xed points would result in the same

membrane performance. The lower ®xed point is the

transmembrane pressure required to establish a Darcy

¯ux equal to the condensation rate. (For all mem-

branes tested this lower ®xed point was below the

sensitivity of the testing equipment used in this study.)

Higher pressures would neither result in a higher ¯ux,

due to a lack of material to transport, nor would higher

pressures result in the retreat of the gas±liquid inter-

face into the membrane. We propose that the gas±

liquid interface (in a porous membrane) would remain

at the air±membrane interface as long as the trans-

membrane pressure is less than the bubble point

pressure, PBP, given by [6]

PBP � �4F�� cosine���=d; (4)

where F is the shape factor (dimensionless), � the

surface tension (N/m), � the water/pore wall contact

angle (8), and d is the diameter of pore (m). Eq. (4) is a

static relationship and establishes the upper limit for

the operational transmembrane pressure.

2.4. Overall mass transfer coefficient

The ®nal conclusion from the prior discussion is

that km, the membrane mass transfer coef®cient, is not

a factor in the overall mass transfer coef®cient, ko. We

must, however, consider mass transfer at the mem-

brane surface. Operational observations con®rmed

that a liquid water ®lm does not form on the humid

air side of the membrane during dehumidi®cation

operations. The bulk membrane material remains

exposed to the humid air. Consider a membrane

cross-section as represented in Fig. 3. Water vapor

impacting the membrane surface can strike either the

solid membrane surface or the water/air meniscus in a

membrane pore. The water molecule's probability of

sticking (the sticking coef®cient) to the surface of the

membrane depends on which surface it strikes.

Furthermore, any water molecule sticking to the solid

surface would have to migrate, via surface diffusion

since a water ®lm does not form, to a pore for transport

through the membrane. Therefore, a reasonable

hypothesis would state that the overall sticking coef®-

cient is related to the ratio of water to solid membrane

surfaces. That is to say, Sw (the sticking coef®cient of

the total membrane surface) is a function of the

membrane porosity, �. Further, viewing kbl as a mea-

sure of the contact rate of water molecules with the

membrane surface and Sw as the probability of the

water molecules remaining on the surface (conden-

sing) the following equation results:

ko � Swkbl; (5)

where Sw is the sticking coef®cient of the total mem-

brane surface, a function of membrane porosity, �(dimensionless), Sw�Sm(1ÿ�)�Sp�, if the rate of

water transport across the membrane surface to a pore

is not a rate-determining step; Sm the sticking coef®-

cient of the bulk membrane material (dimensionless),

Sp is the sticking coef®cient of the water in the

membrane pores (dimensionless). Sw has a positive

functionality with porosity since we assume that Sp is

greater than Sm. Therefore, Eq. (5) predicts that ko will

have a positive scale or functionality with �.

3. Experimental equipment and methods

3.1. Equipment setup

The selection of test equipment focused on the need

to test the application of the dehumidi®cation process

to a speci®c application, the control of humidity in a

plant growth chamber. A closed 25 l container simu-

lated a plant growth chamber for the bench-top experi-

mentation. The varying of the surface area of exposed

water within the container simulated various rates of

plant transpiration. Speci®cally the variation resulted

from using the entire cross-sectional area of the con-

tainer (0.08 m2) to hold water vs. using only a 1 l

beaker (0.008 m2) to hold water. A small fan mounted

inside the container maintained a well-mixed air

condition. An air pump (compressor) removed air

72 P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81

from the container and fed it to the membrane unit

being tested. Air exiting the membrane unit returned

to the container. Therefore, the container behaved like

a CSTR (continuous stirred tank reactor) with relative

humidity and temperature sensors installed within the

container serving two purposes. The ®rst purpose was

to de®ne the input conditions to the membrane unit,

and the second purpose was to measure the steady-

state conditions achievable by the membrane unit

when attached to a plant growth chamber.

The cooling water for the membrane unit followed a

loop as described below and as shown in Fig. 4. The

coolant water loop:

1. Passed through a constant-temperature bath to

achieve the desired temperature.

2. Entered a graduated cylinder through a parafilm

membrane thereby exposing the water to atmos-

pheric pressure. The graduated cylinder allowed

tracking of the condensation rate via water volume

readings.

3. Traveled from the graduated cylinder, through a

restrictor valve, a pressure sensor, the humidity

control membrane, another pressure sensor, and

finally, into the inlet of a gear water pump. Since

the graduated cylinder is at atmospheric pressure,

this arrangement produces a pressure gradient

Fig. 3. Hydrophilic membrane cross-section. Shown are the potential fates of a water molecule striking the solid fraction of the humid air/

membrane interface. The water either rebounds off the surface or adsorbs. Any adsorbed molecule must transport along the surface to a pore

for recovery as condensate; during this transport de-adsorption could occur. Therefore, the overall sticking coefficient of the membrane surface

is a function of the ratio of solid/pore surface areas.

Fig. 4. Flow diagram for experimental setup. Showing the humid air loop (top) and the coolant water loop (bottom); details of membrane

module(s) appear(s) in Fig. 5.

P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81 73

across the membrane in the humidity control unit

with the low pressure on the water side of the

membrane. Adjustment of the restrictor valve pro-

duced variations in the transmembrane pressure.

The transmembrane pressures used were, for mixed

cellulose ester, 6.8 KPa; for metal, 4.1 KPa; and for

ceramic, 1.7 KPa.

4. After flowing through a rotameter the water

returned to the constant temperature bath to com-

plete the loop.

3.2. Membrane modules and membrane materials

Fig. 5 shows the three membrane holding modules

used for the experimentation. The mixed cellulose

ester hollow ®ber membrane module contained

approximately 40 ®bers. The metal and ceramic mem-

brane modules each contained only one tube. In

keeping with the selection of test equipment and

conditions that resemble those in a plant growth

chamber for space ¯ight, the membrane modules

selected represent the size and weight constraints of

space-¯ight hardware. At times, the size limitations

imposed by integration into space shuttle missions

were contrary to the desire to test membranes of the

same surface area. Also, the surface areas within

modules were limited to those available from mem-

brane suppliers. The result was the tested membrane

surface areas appearing in Table 1. Humid air ¯owed

in the tube side of all of the tested membranes while

the cooling water ¯owed counter-current on the shell

side of the membrane module.

The cooling water ¯ow rate was high enough to

allow a constant wall temperature assumption. Water

temperatures were �58C or �108C depending on

experimental design. Due to heat transfer within the

membrane wall, the measured water temperature dif-

fers from the temperature of the membrane surface in

contact with the humid air. The following are the

calculated maximum temperature differences across

the saturated tested membranes, in which the numbers

in parentheses indicate the thermal conductivities,

based on volumetric mixing rules, used for the calcu-

lation

� Mixed cellulose ester (h�0.4 W/m/K) [3],

�T�0.098C.

� Metal (h�12 W/m/K) [7], �T�0.38C.

� Ceramic (h�2 W/m/K) [7], �T�0.88C.

These �T's seem to give a slight mass transfer

advantage to the cellulose ester membrane (a cooler

surface) and a disadvantage to the ceramic membrane

when compared to the metal membrane. However, all

of these �T's are within the margin of error of the

temperature scale readings of the experimental equip-

Fig. 5. Diagram of the three membrane holding modules.

74 P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81

ment (18C�0.58C) and negligible when comparing

mass transfer rates among the different membranes.

All membranes tested had pore sizes of 0.5 mm or

less, primarily to act as a pathogen barrier. The reason-

ing was that the primary application being examined is

dehumidi®cation with condensed water recycle to the

plants. Pathogen control is vital to the health of the

plants and important in earth-based application of

dehumidi®cation units. While 0.2 mm pore size is

preferred for pathogen control, 0.5 mm will remove

the majority of airborne pathogens. Unfortunately, in

the case of porous metal membranes, the smallest pore

size available from our vendor in tubular form was

0.5 mm. Speci®cally, the experimental pore sizes

appear in Table 1 along with the relevant membrane

material and module speci®cations.

4. Results and discussion

4.1. Interpretation of data

The experimental setups described in Section 3

produced the following raw data for a given combina-

tion of exposed water surface area in the container,

coolant water temperature, humid air ¯ow rate through

the membrane module, and membrane module type:

� Steady-state relative humidity achieved in the

chamber as a function of humid air treatment rate

(Fig. 6).

� Entry and exit humidities for the membrane mod-

ule.

� Condensation rates.

For the purposes of prototyping a membrane system

design for space ¯ight only, direct evaluation of the

raw data (Fig. 6) had value, since the selected condi-

tions represented the size and weight constraints

established by space-¯ight requirements. These raw

data and simple derived data, such as membrane ¯ux

vs. driving forces, however, depend not only on the

membrane transport properties but also the humid air

¯uid dynamics within the membrane module. The

humid air ¯uid dynamics effects on mass transfer,

kbl (see Eqs. (1) and (5), depend in part on the

membrane module geometry which, owing to manu-

facturing procedures and experimental necessity, var-

ied among the membrane types tested. The intent of

this article is to de®ne and compare the hydrophilic

dehumidi®cation membrane transport properties and

not the comparison of different membrane module

designs. Therefore, it is appropriate, indeed essential,

to normalize the transport rate data against a common

scale of humid air ¯uid dynamics. To that end, Fig. 7

is a plot of the overall mass transfer coef®cient against

the humid air Reynolds number. The overall mass

transfer coef®cients in Fig. 7 were determined from

the entrance and exit humidities for a given value of

the coolant water temperature. The overall mass

Table 1

Dimensional characteristics of membranes studied

Membrane Pore size

(mm)

Porosity Surface

area (cm2)

Membrane

length (cm)

Fiber or

tube ID (cm)

Fibers or

tubes in module

Manufacturer

Mixed cellulose ester 0.1 0.85 100 13.0 0.06 40 Microgon

Ceramic a-Al2O3 0.2 0.33 50 25.4 0.635 1 US Filter

Metal 316LSS ± sintered 0.5 0.30 30 15.24 0.635 1 Mott

Fig. 6. Achievable steady-state chamber humidity vs. air treatment

rate. Membrane wall temperature�108C (except for metal

membranes with 7.58C). MCE�0.1 mm mixed cellulose ester

membrane, metal�0.5 mm sintered metal (316 SS) membrane,

and ceramic�0.2 mm ceramic membrane.

P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81 75

transfer coef®cients reported here are the log-mean

driving force coef®cient, ko(ln) [8]. The ko(ln) quantities

under the conditions of this study are equivalent to

ko(loc) (Eq. (1)), when we assumed a constant ko(loc)

with membrane length. Fig. 7 is not a precise com-

parison of ko vs. kbl since the membrane aspect ratio,

L/dia, (dia�inner diameter in m and L�membrane

length in m) would also have an impact on the kbl,

particularly for the metal and ceramic tubular mem-

branes of this study.

The best way to normalize the data would be to

compare the measured Sherwood numbers

(Sho�ko dia/D, where D is the diffusivity of water

vapor in air in m2/s) among the different membrane

types, plotted against the predicted boundary layer

Sherwood numbers (Shbl) from the humid air ¯uid

dynamics. Such a plot would help to con®rm Eq. (5).

Fig. 8 is therefore a plot of Sho vs. Shbl, where Shbl was

estimated by adapting the analysis of Kay and London,

1955 [8,9] for short heat exchangers to our short mass

transfer units using the analogy between heat and mass

transfer correlations [8] (see Appendix A). Fig. 8

indicates that a comparison of the data used for

prototyping a membrane system design for our

space-¯ight application cannot be used for generally

con®rming or denying Eq. (5) since the ¯uid dynamic

mass transfer ranges used for mixed cellulose ester

and the other membranes do not adequately overlap.

Due to the order-of-magnitude difference in the mem-

brane inner diameters, we are limited to comparison of

the scaled overall performance among the three mem-

brane compositions.

4.2. Performance comparison of different hydrophilic

membrane materials

Returning to Fig. 7 we see that, when normalized

against Reynolds number, the mixed cellulose ester

membranes exhibit superior performance. This better

performance may be a function of porosity as indi-

cated by Eq. (5) or may re¯ect the much smaller

diameter of the hollow ®ber membranes. Fig. 9, a

plot of ko/� vs. Reynolds number, is an illustration

of the potential role of membrane porosity on overall

mass transfer coef®cient. The porosities, �, of the

tested membranes were: mixed cellulose ester,

��0.85; metal, ��0.30; and ceramic, ��0.33.

Fig. 9 is consistent with Eq. (5); however, it cannot

be taken as de®nitive proof. The divergence in ko for

the mixed cellulose ester when operated at 68C vs.

108C, shown in Fig. 7, has no apparent explanation

except possibly the temperature functionality of Sm or

�, or they are due to experimental artifacts rather than

indications of a membrane mass transfer property.

During the construction of the ®nal space-¯ight

hardware, we had the opportunity to measure the

Fig. 7. Overall mass transfer coefficient, ko, vs. humid air Reynolds number. Comparison of performances for different hydrophilic porous

membranes. MCE�0.1 mm mixed cellulose ester, metal�0.5 mm sintered metal (316 SS), and ceramic�0.2 mm ceramic. For each material two

coolant water temperatures were used as noted. Upper limit of laminar flow (Re�2100) shown.

76 P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81

overall mass transfer coef®cient, ko, of two metal

membranes with approximate porosities of ��0.068

and ��0.26 under identical air/membrane boundary

layer transfer coef®cient, kbl. The membrane with

��0.26 had a ko that was 12% greater than the

membrane with ��0.068. While this difference in

ko is consistent with Eq. (5), this 12% difference

was also within the margin-of-error (90% con®dence

interval) for the measurements and cannot be used as

de®nitive proof of Eq. (5). In summary, all experi-

Fig. 8. Plot of experimental Sho vs. calculated Shbl. This figure indicates that a comparison of the data cannot be used for confirming the

functionality ko on � Eq. (5) since the fluid dynamic mass transfer regions of MCE do not overlap with those of the other membranes. Only

laminar flow data shown. MCE�mixed cellulose ester, and metal�sintered metal (316 SS). For each material two coolant water temperatures

were used as noted in Fig. 7.

Fig. 9. ko/� vs. Re. The overall mass transfer coefficient divided by membrane porosity vs. the humid air Reynolds number. This figure shows

a potential relationship between the overall mass transfer coefficient and the membrane porosity, �. MCE�mixed cellulose ester membrane,

��0.85; metal�sintered metal membrane, ��0.3; ceramic�ceramic membrane, ��0.33. Upper limit of laminar flow (Re�2100) shown. For

each material two coolant water temperatures were used as noted in Fig. 7.

P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81 77

mental observations were consistent with, but not

proof of, the sticking-coef®cient mixing rules used

in Eq. (5).

The operational comparison of hydrophilic mem-

branes evaluated the following four characteristics:

1. Condensation ¯ux per surface area, ko.

2. Ease of start-up.

3. Stability.

4. Durability.

Mixed cellulose ester hollow ®ber membranes have

the highest condensation ¯ux per surface area, the ®rst

characteristic. This may be a function of porosity as

stated in Eq. (5) or just of the hollow ®ber geometry.

The second and third characteristics re¯ect the hydro-

philicity (wettability) of the membrane as measured

by the water/pore wall contact angle, �. A highly

hydrophilic membrane material has � near zero and

easily draws water into the membrane to establish the

water/air interface of Fig. 3. The establishment of this

water/air interface constitutes `̀ Priming'' which is a

required start-up procedure. Mixed cellulose ester

membranes are self-priming while microporous

metal and ceramic membranes require an active

saturation step, as described below in Section 6. Sta-

bility is a function of the range of transmembrane

pressures allowed during operation. Consider that if

the difference between the two pressure limits, �P for

Darcy ¯ow and PBP, the bubble point pressure, is

small, it would be dif®cult to maintain a stable work-

ing membrane. Eq. (4) indicates that the stable pres-

sure range is also a function of the water/pore±wall

contact angle, �. The more wettable or hydrophilic

(small � or high bubble point, PBP) the membrane, the

more stable is its operation. For instance, the bubble

point of mixed cellulose ester tested was 600 KPa

while the bubble point of the metal membrane was

9 KPa. Therefore, mixed cellulose ester membranes

also compare favorably for the second and third

characteristics.

Unfortunately, mixed cellulose ester membranes

were inferior to metal membranes in one signi®cant

category, durability. Dehumidi®cation using mixed

cellulose ester membranes fails after operation

times of only hours to days. Visually, this failure takes

the form of air bleeding across the membrane into the

bulk water phase. The solution of this operational

failure problem is one important area for future

research.

In summary, the optimization of a membrane mate-

rial appears to depend on the following four factors:

1. high porosity, �, (condensation rate);

2. low water contact angle, � (ease of start-up);

3. bubble point, PBP (operational stability); and

4. structural integrity (durability).

5. Discussion of polymer membrane performance

Although the above-mentioned mixed cellulose

ester membrane's operational failure eliminates it

from utilization, the ability to make polymer mem-

branes of much higher porosity than metal or ceramic

membranes [10] creates an incentive for the continued

exploration of the use of polymer membranes for

hydrophilic dehumidi®cation. The advantage of hol-

low ®ber membranes in dehumidi®cation is also sig-

ni®cant. Fig. 10 shows a data set for dehumidi®cation

using mixed cellulose ester hollow ®ber membranes

plotted as the extent of dehumidi®cation vs. reduced

®ber length (length/humid air velocity). The following

equation calculates the extent of dehumidi®cation

from these raw data:

� � �Hin ÿHout�=�Hin ÿHwall�; (6)

where � is the extent of dehumidi®cation, Hin the

absolute humidity of air entering membrane ®ber (kg

of H2O/kg of dry air), Hout the absolute humidity of

air exiting membrane ®ber (kg/kg), and Hwall is the

absolute humidity of air in equilibrium with the cool-

ant water (kg/kg).

Eq. (6) represents the ratio of dehumidi®cation that

occurred vs. the maximum dehumidi®cation that the-

oretically could occur. If ko is known, the extent of

dehumidi®cation could be estimated from

��1ÿexp{ÿ�ko(dia)L/V} (where dia is the ®ber inner

diameter, L the ®ber length, and V is the volumetric

¯ow rate) which indicates the advantage of the small

diameters of the hollow ®ber membranes. Fig. 10

indicates that improvement in a dehumidi®cation

unit's ef®ciency would result from operating the

hollow ®bers within a reduced length of 0.02 s, which

is to say that a module with a series of short ®bers is

much more effective than a module with longer ®bers.

Fig. 10 mainly re¯ects a reduction in the mass transfer

driving force with ®ber length, and therefore, a reduc-

tion in the ef®ciency of the dehumidi®cation unit

78 P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81

when considering membrane surface area. Hollow

®ber polymer membranes would, therefore, allow

the development of a dehumidi®cation `̀ ®lter'' made

of short ®bers (<1 cm long) and having a low air ¯ow

pressure drop. This is consistent with observations

made by researchers in the ®eld of vacuum membrane

distillation [11]. Hollow ®ber technology would also

allow the development of minimal pressure-drop mod-

ules with shell side humid air ¯ow [12].

6. Metal and ceramic membrane priming

As presented above, the dehumidi®cation ef®ciency

of membrane-based systems is directly related to the

porosity of the membrane, or, more precisely, the void

volume fraction containing water which is also the

volumetric water content, �. Saturation, S, and volu-

metric water content, �, are interrelated via porosity,

�:��S� [13]. Due to the high water contact angle of

the metal and ceramic membranes, an operator cannot

rely on passive uptake of water to accomplish a high

saturation. In order to accomplish this, the operator

must actively saturate or `̀ prime'' the membranes.

Saturation or priming techniques generally consist of

the following steps (1) saturate the membrane with a

`̀ priming'' ¯uid and (2) force water through the

membrane via a pressure differential, `̀ priming pres-

sure'' to replace the `̀ priming'' ¯uid with the desired

liquid water. The `̀ Conventional Wisdom'' for water-

saturating a membrane is to use a water-miscible

`̀ priming'' ¯uid with a low surface tension, namely

isopropanol. Isopropanol cannot be used in an

enclosed plant growth chamber on a spacecraft; there-

fore, we avoided isopropanol and used carbon dioxide

(CO2) gas. Carbon dioxide gas, as the `̀ priming'' ¯uid,

displaces the air from the membrane pores. Due to the

high solubility of carbon dioxide in water, the water

forced through the membrane during step (2) now

dissolves the carbon dioxide replacing the pore spaces

with liquid water. In addition, we found that carbon

dioxide (CO2) gas had superior `̀ priming'' ¯uid qua-

lities when compared to isopropanol.

7. Conclusions

In studying hydrophilic membrane-based humidity

control, we de®ned and proposed the membrane prop-

erties that determine performance. In summary, the

following are the critical factors for choosing a hydro-

philic membrane material:

1. High porosity, � (in theory, for maximum

condensation rate).

2. Low water contact angle, � (for ease of start-up).

3. High bubble point, PBP (for operational stability).

Fig. 10. Experimental extent of condensation (108C coolant water) vs. reduced fiber length (length/velocity, s). For mixed cellulose ester

hollow fiber membrane; Hi�absolute humidity at `̀ i'', `̀ in''�entrance to fiber, `̀ out''�exit of fiber, and `̀ wall''�fiber wall.

P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81 79

4. Reliable structural integrity (for durability).

Due to the current state-of-the-art in membrane

manufacturing, polymer membranes can have higher

porosities than metal or ceramic membranes. With this

factor in mind, polymer membranes, as represented by

mixed cellulose ester, would be the superior mem-

brane material if the problem of operational failure

(see Section 4) could be solved.

Hollow ®ber dehumidi®cation membrane modules

would have a high ef®ciency and low pressure drops.

However, the mixed cellulose ester membranes tested

have impractical short operational lives. Therefore,

the current state of the art for membrane-based dehu-

midi®cation involves using rigid microporous materi-

als such as metal membranes. These rigid microporous

materials require a priming step during system start-up

to achieve optimal performance. This priming step is a

membrane saturation process.

8. List of symbols

ko overall mass transfer coefficient (m/s)

ko(loc) local overall mass transfer coefficient (m/s)

ko(ln) log-mean overall mass transfer coefficient

(m/s)

kbl air/membrane boundary layer mass transfer

coefficient (m/s)

km membrane mass transfer coefficient (m/s)

Sw sticking coefficient of the total membrane

surface (dimensionless)

Sm sticking coefficient of the bulk membrane

material (dimensionless)

Sp sticking coefficient of the water in the

membrane pores (dimensionless)

h thermal conductivity, (W/m/K)

j molar water flux (mol/m2/s)

J water flux (m/s)

K membrane permeability (m/s Pa)

R ideal gas constant (m3 Pa/mol K)

T absolute temperature (K)

P air pressure (Pa)

�P transmembrane pressure (Pa)

PBP bubble point pressure (Pa)

X wetted membrane thickness (m)

F shape factor (dimensionless)

� surface tension (N/m)

� water/pore wall contact angle (8)

d diameter of pore (m)

� porosity (m3/m3)

� volumetric water content (m3/m3)

S saturation (dimensionless)

Hi absolute humidity at location `̀ i'' (kg of

water/kg of dry air)

Ha absolute humidity of the bulk air (kg/kg)

Hw absolute humidity of air in equilibrium with

the liquid water (kg/kg)

Hin absolute humidity of air entering membrane

fiber (kg/kg)

Hout absolute humidity of air exiting membrane

fiber (kg/kg)

Hwall absolute humidity of air in equilibrium with

the coolant water (kg/kg)

� extent of dehumidification (dimensionless)

D diffusivity of water vapor in air (m2/s)

dia inner diameter of tubular membrane (m)

L membrane length (m)

V volumetric flow rate (m3/s)

Acknowledgements

The authors greatly appreciate The National

Science Foundation's (NSF) Research Experience

for Undergraduates Program for supporting the efforts

of Mr. Jedrick Burgos. Our colleagues in the NSF

Center for Separations Using Thin Films, University

of Colorado, supplied expert advice. We also thank the

National Aeronautics and Space Administration for

supporting this work under Grant NAGW-1197 to the

BioServe Space Technologies Center at the University

of Colorado, Boulder.

Appendix A

The following equation estimates Shbl for

1<Re(Sc)dia/L<100:

Shbl � 3:5�Re�Sc�dia=L�0:2; (A.1)

where Re is the Reynolds number of humid air ¯ow

(dimensionless), Sc the Schmidt number of humid air

¯ow (dimensionless), dia the inner diameter of tube

(m), and L is the length of membrane tube (m).

Eq. (A.1) was adopted from Kay and London, 1955

[8,9] using the analogy between heat and mass transfer

correlations [8] by ®tting a curve over the correlation

80 P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81

for 1<Re(Sc)dia/L<100. Eq. (A.1) takes into account

that signi®cant mass transfer in the ceramic and metal

membranes occurs prior to fully developed ¯ow (i.e.

low aspect ratios, L/dia). As the aspect ratio (L/dia) in

Eq. (A.1) goes to in®nite (i.e. hollow ®ber mem-

branes) the solution of Kay and London becomes

Shbl�3.657 which is consistent with the solution of

Davis and Parkinson, 1970, for mass transfer inside a

hollow ®ber membrane of Shbl�4.0 [6].

References

[1] A. Hoehn, D.J. Chamberlain, S.W. Forsyth, K. Gifford, D.S

Hanna, M.B. Horner, P. Scovazzo, J. Smith, L.S. Sodieck,

P.W. Todd, Plant generic bioprocessing apparatus: a plant

growth facility for space flight biotechnology research,

Proceedings of the Sixth European Symposium on Life

Sciences Research in Space, Trondheim, Norway, June 1996.

[2] N.A. Duffie, R.J. Bula, R.C. Morrow, P.E. Quirin, P.D.

Macaulay, S.R. Wranovsky, J.C. Vignali, T.W. Tibbits,

Humidity and temperature control in the AstrocultureTM

flight experiment, SAE 24th International Conference on

Environmental Systems and 5th European Symposium on Space

Environmental Systems, Friedrichshafen, Germany, SAE Tech-

nical Paper 941282, Warrendale, PA, 20±23 June 1994.

[3] D.D. Newbold, Analysis of a membrane-based condensate

recovery heat exchanger (CRX), Bend Research, Inc., SAE

Technical Paper Series, 23rd International Conference on

Environmental Systems, Colorado Springs, CO, SAE Tech-

nical Paper 932089, Warrendale, PA, 12±15 July 1993.

[4] G. Noyes, Microporous hydrophobic hollow fiber modules for

gas±liquid phase separation in microgravity, SAE 23rd

International Conference on Environmental Systems, Color-

ado Springs, CO, SAE Technical Paper 932090, Warrendale,

PA, 12±15 July 1993.

[5] R.J. Ray, A membrane-based subsystem for water±vapor

recovery from plant growth chamber, NASA Contractor

Report 177602, September 1992.

[6] W.S.W. Ho, K.K. Sirkar (Eds.), Membrane Handbook, Van

Nostrand Reinhold, New York, 1992.

[7] R.H. Perry, D.W. Green, J.O. Maloney, Perrys Chemical

Engineering Handbook, 6th ed., McGraw-Hill, New York,

1984.

[8] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenom-

ena, Wiley, New York, 1960.

[9] W.M. Kays, A.L. London, Compact Heat Exchangers: A

Summary of Basic Heat Transfer and Flow Friction Design

Data, The National Press, Palo Alto, CA, 1955.

[10] M. Mulder, Basic Principles of Membrane Technology, 2nd

ed., Kluwer Academic Publishers, Dordrecht, Netherlands,

1996.

[11] S. Bandini, C. Gostoli, G.C. Sarti, Separation efficiency in

vacuum membrane distillation, J. Membr. Sci. 73 (1992) 217±

229.

[12] D.D. Newbold, S.B. McCray, D.L. Millard, R. Ray,

Performance of a membrane-based condensate-recovery heat

exchanger, Bend Research, Inc., SAE Technical Paper Series,

26th International Conference on Environmental Systems,

Monterey, CA, SAE Technical Paper 961356, Warrendale,

PA, 8±11 July 1996.

[13] A.T. Corey, Mechanics of Immiscible Fluids in Porous Media,

Water Resources Publications, Highlands Ranch, CO, 1994.

P. Scovazzo et al. / Journal of Membrane Science 149 (1998) 69±81 81